Sports Coaching App

Client Objectives

Our client was working in a sports broadcast company in India & managing the broadcast of the most popular sport cricket, during his job experience he has covered a lot of top coaching centres & national sports persons on playing technique. He is bullish on the potential of AI and invested in making an AI-based coaching platform. He is embracing the mission to overcome the problem of efficient coaching. The current problem involves youth sports have become an expensive proposition, costing families thousands of dollars each year & studies show that 70% of kids stop playing sports by the age of 13 due to cost and lack of access to quality coaching. We at FiftyFive Technologies partnered with our client to build a world-class coaching platform & offer every young athlete access to the same coaching programs and extensive biomechanical analysis used by the best athletes in the world, and the same personalized training protocols.

Key Solutions

Our skilled & self-propelled tech resources (Data Scientists, Frontend engineers & QA engineers having 10, 8 & 10 years of experience respectively) joined the client’s shared team to build solutions in AI & ML

• Our experts build AI models by leveraging Computer vision and image processing & IoT for Cricket sport.
• Developed AI-based engine which analyzes the body key joints & recognizes the player’s body moments & gives suggestions to improve the player’s performance
• Built Machine learning pipeline on GCP, utilised PyTorch & OpenCV.
• Built the model that can extract the 133 body key joints of the subject player’s body, analyze the subject player’s recording and provide feedback on the improvement areas. The system is also advanced to extract and track only the subject player from the entire team
• Worked on Object detection (person & ball detection), Object Tracking & Pose estimation, SVM, LSTM
• We successfully ran POCs and demonstrated the model to the client.
• Built mobile & web apps where users can upload videos and take feedback from AI models.
• Overall tech stacks utilized are- Python, Pytorch, MMCV, Time Series Analysis, Kubeflow, GCP, Flutter, Firebase
• The Model achieved an accuracy score of around 90%